General Information

The process of collecting analytics consists of sending events and saving them in the database. The table in the database has the following structure:

• event_id - unique event identifier;

• event_name - type of event;

• timestamp - date/time of the event;

• session_key - Garderobo session identifier;

• user_id - unique user identifier;

• page_source - type of page from which the transition to the product card or the addition to the cart from a popup occurred;

• widget_source - type of widget through which the transition to the product card or the addition to the cart from a popup occurred;

• target_id - event target identifier, with which actions were performed (product/category/look ID, etc.);

• price - product price;

• device_type - type of device (mobile/tablet/desktop).

Event types

Event	Feature	Description
widget_click	page_source; widget_source; target_id;	Triggered when user navigates to a product description page through a widget, as well as before they add an item to cart from a popup.
add_to_cart	target_id;	Triggered when the product is added to cart via the form on the PDP (Product Detail Page) or via the widget popup.
purchase	target_id; transaction_id;	This is triggered after the checkout process is completed. If an order contains multiple identical items, each purchase is registered as a separate event with the same transaction_id.
open_category	target_id;	Triggered when user navigates to the section with the list of products within a category.
open_look_popup	target_id;	Triggered when the user clicks on a look and opens a popup window.
change_popup_item	target_id;	Triggered when the user replaces a product in the look within the popup window.

The same data is also used to train the recommendation model to improve the quality of recommendations.

Data generation for reports

To generate e-commerce reports, it is necessary to identify the source which led the user to interact with an item before this item is added to cart and purchased. For example, to determine revenue generated through widgets (sum of prices from 'purchase' events), we check for an earlier 'product_click' event with the specified 'page_source' and 'widget_source' for the current user.

This event logging approach takes cross-platform purchases into account. For example, if an item was added to 'Favorites' through a widget on a mobile phone and later purchased from a pc by the same user, we will be able to track this user's full history of interaction with this item.

Orders

In this section, the data is grouped into two categories:

• Without Widgets - all orders where no items were purchased through widgets;

• Mixed - all orders that include at least one item purchased through widgets.

Available data:

• AOV - Average Order Value;

• UPT - Average number of units per order;

• Purchased Items - Total number of items purchased;

• Transactions - Total number of transactions;

• Revenue - Total revenue. To assess the effectiveness of different types of widgets, a table was added to the report:

Widget Statistics

This table displays types of widgets, types of pages on which the widgets are displayed, the number of clicks on the widget, the conversion rate (CR) from click to purchase, the number of items purchased through widget, and the total revenue generated from items purchased through each widget type.